Studies have suggested that more carbon is fixed due to a large increase in photosynthesis in plant–soil systems exposed to elevated CO2 than could subsequently be found in plant biomass and soils –- the locally missing carbon phenomenon. To further understand this phenomenon, an experiment was carried out using EcoCELLs which are open-flow, mass-balance systems at the mesocosm scale. Naturally occurring 13C tracers were also used to separately measure plant-derived carbon and soil-derived carbon. The experiment included two EcoCELLs, one under ambient atmospheric CO2 and the other under elevated CO2 (ambient plus 350 μL L− 1). By matching carbon fluxes with carbon pools, the issue of locally missing carbon was investigated. Flux-based net primary production (NPPf) was similar to pool-based primary production (NPPp) under ambient CO2, and the discrepancy between the two carbon budgets (12 g C m− 2, or 4% of NPPf) was less than measurement errors. Therefore, virtually all carbon entering the system under ambient CO2 was accounted for at the end of the experiment. Under elevated CO2, however, the amount of NPPf was much higher than NPPp, resulting in missing carbon of approximately 80 g C m− 2 or 19% of NPPf which was much higher than measurement errors. This was additional to the 96% increase in rhizosphere respiration and the 50% increase in root growth, two important components of locally missing carbon. The mystery of locally missing carbon under elevated CO2 remains to be further investigated. Volatile organic carbon, carbon loss due to root washing, and measurement errors are discussed as some of the potential contributing factors.